Pipe cap filters for 10GHz

The most common problem in a transverter system is local oscillator leakage, this is the local oscillator signal get through the TX mixer to the power amplifier chain, becoming present at the antenna connector at some level. LO leakage is usually a sign of poor design, but it can occurs also by defective or wrong shielding or even by a metallic enclosure box.

In my case the problem was the metallic enclosure used for the transverter. With the cover removed everything was fine but once the cover was put in its place, the local oscillator was present at the output at nearly 5mW of power: Just unacceptable.

The problem was the cover acts as a mirror, making the radiation from the stripline filter at the LO multiplier to be reflected towards the strip line filter in the TX converter, both working at the same frequency: 9936 MHz. As you can see, unshielded striplines are not a good idea, as they acts as antennas.

My first try was to add some microwave absorptive material but I only got modest results. Then the most obvious solution would be to shield at least one of the stripline filters, but I was afraid to detune them. So I opted for another solution: A pipe cap based bandpass filter.

My reference was this article from W1GHZ. After reading it I have a clear idea of how a pipe cap filter works. W1GHZ uses 1/2" pipe caps for 10 GHz , so I bought some 12mm pipe caps, as they are the closest ones to 1/2"

But soon I discovered I was wrong. The 1/2" refers to the internal diameter, and my pipe caps have 12mm of external diameter. They were smaller than the one used for 10 GHz, so I ended with some 16mm pipe caps, who have 14mm of internal diameter. The rest of the dimensions were the ones at W1GHZ's article.

Because I need high selectivity in the filter I decided to use two pipecaps in series. This is not a recommended option, because both filters adjustment interacts between them. Anyway, because a 10GHz transverter is almost a single frequency device (10368.2 MHz) I decided to go ahead with this option.

The first step is to drill a hole in the top of the pipe cap and solder a nut there. This is easy. Then you will need a small piece of PCB to solder the caps. First I shaped the semirigid cables to accommodate them to the final position of the filter, just at the power amplifier's input. Once all was in position, I soldered the coaxial cables to the PCB. The next step was to decide the probe length.

A larger probe will produce low loses, but also wider bandwidth. Shorter probes produce higher loses, but narrower bandwidths. Because my Eyal Gal module has a lot of gain, I was able to sacrifice loses to get a sharper filter, so finally I cut the probes at 2.8mm, this is 0.3mm less than 1/8", the minimum stated at W1GHZ's article.

The next step is the most difficult one in the construction of a pipe cap filter: When you solder the cap to the PCB. You need a high power solder iron, and you need to solder it as quick as possible, otherwise all previous solders will melt, ruining your pipecap. I put the caps in place with the aid of some wires and I used a 100W solder iron.

Once the filter is finished, it's time to tune it. In a pipe cap filter, the screw acts as a quarter wave length antenna that couples energy between probes. So, to peak a filter at 10368.2 MHz and not at 9936 MHz, it is best to start with the screw all outside the cap and turn it down (into) the cap.

Then, I began to turn the screws slowly down, until I saw the first response at 10368 MHz. Then it was easy: Just to peak to maximum signal while alternatively playing with both screws. It was very easy and the peak was easily identifiable. Impossible to make a mistake. Then I adjusted the transverter's power amplifier gain to get again 1 watt output. Only a slightly readjustment was necessary.

What about the local oscillator leakage? Well, it went from about 5mW (+7dBm) down to nearly -40 dBm, producing a total attenuation of almost 70dBc, pretty nice!

Pipe caps filters are easy to construct with readily available (and cheap) materials. They can have significant in line losses, but if you have gain margin (or can add some MMICs to your project) they are a nice solution.